EP1865654A1 - AES encryption circuitry with CCM - Google Patents

Abstract

The invention concerns circuitry for encrypting at least a part of an input data (Bi) flow and generating a tag based on said input data flow with a same ciphering algorithm and a same key comprising: a first ciphering branch arranged to encrypt said at least part of said input data; a second ciphering branch arranged to generate said tag; and a single key schedule unit arranged to receive said key, to generate at least one sub-key based on said key and to provide said at least one sub-key to said first and second ciphering branches.

Description

FIELD OF THE INVENTION
• The present invention relates to the field of data encryption using a symmetric algorithm, more particularly the Advanced Encryption Standard (AES) algorithm. The present invention relates to a circuit and method of encryption in a combined counter and CBC-MAC mode (CCM).
• The present invention applies to secured data transmission, more particularly to secured wireless networks.
BACKGROUND OF THE INVENTION
• Network security related to wireless local area networks (WLANs), for example according to 802.11 (n or i) standards, wireless USB etc., generally requires encryption according to the Advanced Encryption Standard (AES), implemented in CCM mode. The CCM mode provides both privacy, as well as data integrity. To achieve data privacy, data is encrypted using a secret key known to the transmitting and receiving network nodes. Data integrity is ensured by generating a tag (message authentication code - MAC) based on the data to be transmitted, and then recalculating the tag at the receiving end to ensure that data has not been corrupted.
• Messages to be transmitted are in the form of data packets. Each packet comprises a header giving information such as the destination address of the data and additional authenticated data which can be used for example for authentication of packet header, and a payload. The packet header and any additional authenticated data (grouped hereafter under the name header) are not encrypted as they are used for routing the packet to its destination in the network. However, the header is taken into account to compute the tag. The payload is both encrypted and used to generate the tag, which is also usually encrypted.
• Figure 1A is a block diagram illustrating a conventional example of aprocess 100 for encrypting data packets in counter mode. Header and payload are organized in groups (blocks) of bits the size of which depends on the processing granularity of the process. The example of figure 1 illustrates the case when there are four payload data blocks of plain text P₁ to P₄ to be processed, however the process can be expanded to process any required number of data blocks. Each data block P₁ to P₄ is combined (XOR gates 112, 114, 116 and 118) with an encryption sequence S₁ to S₄ to produce four blocks of cipher text C₁ to C₄. Sequences S₁ to S₄ are generated by encrypting (steps 122, 124, 126 and 128 - algorithm E) different nonce values Nj to N₄ with the same secret key K. In counter mode, the successive nonce values N₂ to N₄ are obtained by incrementing (steps 134, 136 and 138) a first nonce N₁ corresponding to an initialization value IV (for example, zero) of a counter. Encryption in counter mode is preferred to an encryption in cipher block chaining as the algorithm E can be applied (calculated) before receiving the data block.
• In a hardware implementation of the encryption to which the present invention applies, a single unit (logic computation core) implementing the algorithm E is successively loaded with a result value of a counter incremented for each new data block P₁ to P₄ and the result provided by this unit is linked to a first input of an XOR gate the second input of which receives the current data block. At each data block, the key K is provided to the unit.
• Figure 1B is a block diagram illustrating a conventional example of aprocess 200 for computing, according to a cipher block chaining (CBC) method, data integrity data in the form of a tag. For data integrity, both the header and the payload have to be processed. For simplicity, it is assumed that a data block P₀ is the header of a packet comprising payload data blocks P₁ to P₄. Each data block P₁ to P₄ is combined (XOR gates 212, 214, 216 and 218) with the result of the application of a same encryption algorithm E (steps 222, 224, 226 and 228) to the former data block. A first data block (here, the header P₀) is combined (XOR gate 210), before encryption (block 220), with an initialization value IV' (for example zero). The result of the encryption (output of step 228) of the last data block P₄ of the packet provides the message authentication code or tag. The ciphering key K is the same for each computation of the algorithm E.
• For implementing the counter mode, a single logic core computing the algorithm E is used for a hardware implementation of the CBC-MAC. This single core is successively loaded by the output of a circuit forming an XOR gate, a first input of which successively receives the initialization value IV' and the successive results of the algorithm when a second input receives the successive data blocks P₀ to P₄. Again, at each data block, the key K is provided to the circuit.
• Fig. 2 illustrates in a schematic form a conventional example of a process for combining the counter mode and the CBC-MAC computation in order to provide both privacy and integrity.
• Assuming a message (in the form of data packet) comprising r+1 data blocks B₀ to B_r including h+1 blocks B₀ to B_h representing the packet header (and additional authenticated data), which are not to be encrypted and r-h payload data blocks B_h+1 to B_r to be encrypted. All the blocks are processed according to the process 200 (CBC-MAC) of Fig. 1B to generate a tag. The payload data blocks are processed a second time according to thecounter mode process 100 of Fig. 1A to obtain ciphered blocks CB_h+1 to CB_r. The first h+1 blocks B₀ to B_h are sent over the network with the r-h ciphered blocks CB_h+1 to CB_r and the tag C_TAG (usually ciphered using the counter mode).
• By recalculating the tag at the destination based on the decrypted blocks, and comparing this to the transmitted tag, the data integrity of the received packet can be checked. The key K and the initialization values IV and IV' have to be known by the receiver. Therefore choosing zero for the IV's avoids the need to transmit them.
• According to the method described above, every part of the payload of a message to be transmitted is processed twice, once for encryption and a second time for data integrity. Known hardware implementations provide a single computing core, surrounded by suitable logic and registers such that it can be used once for encryption and then for data integrity. Whilst the header of each packet need only be processed for the generation of the tag, it is the payload of the packet that forms the majority of the data in each packet, and thus the throughput is limited by two full processing cycles of the payload of each packet. This solution is thus disadvantageous in that it is slow and inefficient at performing the required algorithm.
• This drawback is particularly present for encryption algorithms using a key schedule (for example, the AES algorithm), i.e. according to which, for each block to be processed, sub-keys are generated from a key K and are successively used in rounds of an iterative process. In such algorithms, computation time of the CCM mode encryption can be critical for the rate of transmission of the data.
• An example of the known method described above applied to the AES is disclosed in "FPGA Implementation AES for CCM Mode Encryption Using Xilinx Spartan-II" - Khoa Vu, David Zier - ECE 679, Advanced Cryptography, Oregon State University, Spring 2003.
SUMMARY OF THE INVENTION
• The present invention aims to at least partially address the above problems discussed in relation to the prior art.
• The present invention more specifically aims at providing a fast unit for both encrypting and/or decrypting data for privacy and generating and/or verifying a tag for integrity, in a combined counter and CBC-MAC mode (CCM mode).
• According to a first aspect of the present invention there is provided circuitry for encrypting at least a part of an input data (Bi) flow and generating a tag based on said input data flow with a same ciphering algorithm and a same key comprising: a first ciphering branch arranged to encrypt said at least part of said input data; a second ciphering branch arranged to generate said tag; and a single key schedule unit arranged to receive said key, to generate at least one sub-key based on said key and to provide said at least one sub-key to said first and second ciphering branches.
• According to a second aspect of the present invention there is provided circuitry for decrypting at least a part of an input encrypted data flow and recalculating a tag based on input data with a same algorithm and a same key comprising: a first deciphering branch arranged to decrypt said at least part of input data; a second ciphering branch arranged to generate said tag; a comparison unit to compare said generated tag with a received tag; and a single key schedule unit arranged to receive said key, to generate at least one sub-key based on said key and to provide said at least one sub-key to said first and second branches.
BRIEF DESCRIPTION OF THE DRAWINGS
• The foregoing and other purposes, features, aspects and advantages of the invention will become apparent from the following detailed description of a number of embodiments, which is given by way of illustration only without limiting the invention, and throughout which reference is made to the accompanying drawings in which:
• Figure 1A (already described) illustrates in schematic form a known process for encrypting data blocks;
• Figure 1B (already described) illustrates in schematic form a known process for generating a tag;
• Figure 2 (already described) illustrates in schematic form a known process for encrypting data and generating a tag;
• Figure 3 illustrates in schematic form circuitry for encrypting data and generating a tag according to a first embodiment;
• Figure 4 illustrates in schematic form the key stages in an encryption algorithm according to the Advanced Encryption Standard;
• Figure 5 illustrates in schematic form circuitry for decrypting data and generating a tag according to the first embodiment;
• Figure 6 illustrates in schematic form pipelined circuitry for encrypting data and generating a tag according to a second embodiment;
• Figure 7 illustrates in schematic form pipelined circuitry for decrypting data and recalculating a tag according to the second embodiment; and
• Figure 8 illustrates in schematic form alternative pipelined circuitry for encrypting data and generating a tag according to the second embodiment.
• For reasons of clarity, only steps and elements which are useful for the understanding of the present invention have been shown and will be described. In particular details of the steps of each round of the AES algorithm and of the key schedule have not been described, the present invention being compatible with any conventional application of the AES algorithm.
DETAILED DESCRIPTION OF THE INVENTION
• The present invention will be described in connection with an application to an AES algorithm. However, it applies more generally to any symmetric algorithm using a key schedule, such as for example DES, RC5 (Rivest cipher 5) or RC6, i.e. the operation of which can be shared in a data part processing the plain text (ciphering) or the cipher text (deciphering) and a key part extracting sub-keys from a (secret) key to be used in rounds of the data part. The simplest key schedule is a duplication of key. However, the key schedule preferably generates a function of the secret key, different for each round.
• In such algorithms the key schedule usually occupies a significant area in a hardware implementation. The present invention takes benefit from the fact that the same key is used for both computing the counter mode and the cipher block chaining message authentication code.
• Figure 3 illustrates in schematic form aprocessing unit 300 for performing both encryption for privacy and tag generation for integrity according to a first embodiment.
• As shown in figure 3, processingunit 300 comprises a firstround logic unit 302 for generating a tag according to CBC-MAC (Cipher Block Chaining Message Authentication Code), a secondround logic unit 304 providing counter mode functionality for encrypting packet payloads, and a commonkey schedule unit 305 for generating sub-keys to be provided to both the first and second, preferably identical,round logic units 302 and 304.
• Theunit 300 receives blocks B_i (i taking successively thevalues 0 to r, where r+1 represents the total number of data blocks) to be processed oninput line 306.Input line 306 and the other lines described herein below are in fact parallel data lines each providing a communication route for a number of data bytes or bits in parallel.Input line 306 is connected to a first exclusive OR (XOR)gate 308 which also receives an input online 310 from atemporary tag register 312.XOR gate 308 performs the XOR function on the signals onlines 306 and 310 and provides the result online 313 to the firstround logic unit 302. The firstround logic unit 302 performs the encryption algorithm (for example AES) on this data, and provides the result online 314 to thetemporary tag register 312, which stores the result before providing the result back toXOR gate 308 online 310 ready to be added bit by bit to the next input data block.Elements 308, 302 and 312 perform a CBC-MAC computation branch reusing thesame unit 302 for each computation of the encryption algorithm. The initialization value is, for example, given by the initial contents (preferably zero) of theregister 312.
• Acounter 316 is provided comprising an input for receiving an initialization value IV (for example, zero) online 318 and a reset input for receiving an input signal online 320. The counter adds 1 on every full cycle of the secondround logic unit 304, and the resulting counter value N_j (j taking successively and arbitrarily thevalues 0 and 1 to r-h-1, where h+1 represents the number of header data blocks) is provided online 322 to the secondround logic unit 304.Unit 304 performs the encryption algorithm, which is the same encryption algorithm as performed by the firstround logic unit 302, and provides the result online 324 which is connected to asecond XOR gate 326.Unit 304 computes encryption sequences Sj in counter mode to be combined with the payload data blocks Bi. Thesecond XOR gate 326 also receives the data block Bi (h+1≤i≤r) online 327. Thesecond XOR gate 326 provides an output online 328 which is the encrypted data block CB_i. In particular, encryption sequence S₁ is used to encrypt the data block B_h+1, the encryption sequence S₂ is used to encrypt data block B_h+2 etc.
• Elements 316, 304 and 326 perform a counter mode computation branch reusing thesame unit 304 for each computation of the encryption algorithm.
• Thekey schedule unit 305 receives asecret key 334, which is known to both the transmitting and receiving parts (for example the nodes of a communication network). Based on thesecret key 334, thekey schedule block 305 generates a series of sub keys K₁, (1 taking successively thevalues 0 to n, where n represents the number of rounds of the ciphering algorithm) which are provided both to the firstround logic unit 302 online 336 and to the secondround logic unit 304 online 338.
• The elements ofprocessing unit 300 also receive synchronization and control signals not shown in figure 3 for controlling the operation of the circuitry as will now be described.Processing unit 300 is preferably a logic array, and can be programmable (FPGA) or not.
• The first and second round logic units process in parallel the data blocks Bi for generating the tag and the nonce values Nj generated bycounter 316 for encrypting the data blocks, but the n+1 sub-keys for each computation of the AES (for each block) are generated by the sharedkey schedule unit 305. For blocks B₀ to B_h, theunit 304 is not activated. The data blocks B₀ to B_h, which are not encrypted, are output online 329 directly from theinput line 306 and are processed in parallel by theunit 302. When the B_h+1^th data block arrives atinput line 306, bothunits 302 and 304 are active and the IV is also provided oninput line 318. The data blocks B_h+1 to B_r are processed by theunit 302 and in parallel (using the same sub-keys) the count values N_j are ciphered byunit 304 which generates the encryption sequences S_j for encryption of the data blocks at theXOR gate 326. Once the B_r^th data block has been encrypted and processed by the firstround logic unit 302, the tag at theoutput 314 is the final tag to be transmitted.
• Preferably, an initial encryption sequence S₀ is used to encrypt the tag. The content of the TEMP TAG register 312 is provided to athird XOR gate 330 for encrypting the tag. In a final iteration of thesecond unit 304 alone, thecounter 316 is reset (line 320), and the second round logic unit performs the algorithm on the initialization value (first nonce N₀). The result is provided online 324 to thethird XOR gate 330. The XOR function is applied to this value and the tag from theTEMP TAG register 312, and the output online 332 is the encrypted tag C_TAG which can be sent with the encrypted data CB_i (h+1≤i≤r) and the packet header B_i (0≤i≤h).
• An advantage of using the initial sequence S₀ for encrypting the tag is that tag encryption can be performed irrespectively of the number r-h of data blocks that are to be encrypted. In alternative embodiments, rather than resetting the counter to generate to the initial encryption sequence S₀ for encrypting the tag, the encryption sequence S₀ is generated during processing of the data block B_h-1 and is stored in a register. It can then be reused for encrypting the tag.
• In the example of the AES, the encryption and tag generation performed by the first andsecond unit 302, 304 both use a series of either 11, 13 or 15 sub-keys of 128 bits, generated by thekey schedule circuit 305, the number of rounds computed by eachunit 302 or 304 being 11, 13 or 15 respectively. Each block of data (circuit 302) or nonce (circuit 304) to be processed is 128 bits. According to some alternative embodiments the initial key addition is included in the first round and then the encryption is executed in 10, 12 or 14 clock cycles.
• Figure 4 is a block diagram illustrating the operation of identicalround logic units 302 and 304 for the AES. This encryption algorithm is described in more detail in the publication "The Design of Rijndael" by Joan Daemen and Vincent Rijmen, available from Springer-Verlag (ISBN 3-540-42580-2) and in the AES standard (FIPS PUB 197).
• The algorithm encodes blocks of plain text (data or nonce) each comprising 128 bits into an encoded block of the same size. Each step in the algorithm processes a matrix of four rows and four columns, each element in the matrix being a byte such that the matrix represents a total of 128 bits.
• The process starts with an initial state 401 (STATE INIT) of the four by four data matrix PT to be encrypted.
• The first step 402 (ADDROUNDKEY) comprises performing an XOR operation bit by bit on each of the elements of the matrix with the first sub-key K₀ to obtain a first intermediate state IS₀. The sub-key is generated by the key schedule unit as described above.
• The second step in the process comprises performing a number n-1 of rounds of the same transformation T, providing an intermediate state IS₁ by processing the previous intermediate state IS₁-₁ and a current sub-key K_l. The number n-1 of rounds required is the number of sub-keys generated minus 2. If the secret key is 128 bits, eleven sub-keys are generated, and therefore this transformation is repeated nine times. Each transformation T comprises four operations.
• A first operation 403 (SUBBYTES) comprises a nonlinear transformation in which each byte of the matrix of the previous state IS_l-1 is replaced by its image in a pre-calculated substitution table, known as an S-box. The S-box can be obtained by the combination of two transformations, a first which inverses each byte of the matrix in the limited field of size 2⁸ (to correspond to a byte), each byte being replaced by its own image. This inversion is followed by an affine transformation. Since 128 bits are encrypted every round, sixteen S-boxes are required in total.
• The second operation 404 (SHIFTROWS) comprises performing a rotation on the three bottom rows of the matrix resulting from theprevious operation 403, and rotates the second row by one byte, the third row by two bytes, and the fourth row by three bytes. The first and second operations can be inverted.
• A third operation 405 (MIXCOLUMNS) comprises considering each column of the matrix resulting from the previous operation as a polynomial within the limited field of size 2⁸, and multiplying each of these polynomials by a combination polynomial P[X] modulo a polynomial M[X].
• A fourth and final operation 406 (ADDROUNDKEY) comprises applying the current sub-key K_l by performing the XOR function bit by bit with each bit of the sub-key and each bit of each byte of the matrix resulting from theprevious operation 405. This operation is the same as theoperation 402, but with a different sub-key and provides intermediate state IS₁. At the end of the n-1^th round,operation 406 provides intermediate state IS_n-1.
• Next, a transformation T', forming the last round, processes the intermediate state IS_n-1 and last sub-key K_n. Transformation T' performs successivelySUBBYTE operation 407,SHIFTROWS operation 408 andADDROUNDKEY operation 409 using the last sub-key K_n. Hence, transformation T' corresponds to transformation T without the MIXCOLUMNS operation.
• If needed, a last step 410 (RESULT FORM) puts the result (ciphered text CT) into the appropriated form to be outputted (online 314 forunit 302 and online 324 for unit 304).
• Whilst figure 4 illustrates multiple identical steps, for example threeADDROUNDKEY steps 402, 406 and 409, in practice only one function unit need be provided for performing each operation. Thus a total of four function units are required for the four operations, and the matrix can be passed through each of these units the required number of times, the appropriate function unit being activated on each cycle. Furthermore, inputs and/or outputs of the operations are usually buffered for synchronization purposes.
• The sub-keys, a different one of each round of the encryption algorithm, are generated by the key schedule unit. The key schedule uses S-boxes to generate the sub-keys from the key K. Four S-boxes are used for generating the n+1 sub-keys and are either to be stored or computed for each data block.
• Figure 5 illustrates in schematic form aprocessing unit 500 for decrypting encrypted messages at the destination node according to the first embodiment. Thecircuitry 500 comprises a firstround logic unit 502 for implementing CBC-MAC, a secondround logic unit 504 for implementing counter mode, and akey schedule unit 506 for generating sub-keys for use by the first and second round logic units. Theprocessing unit 500 receives on aninput line 508 data blocks B_i, for i between 0 and h, which represent the header (and additional authenticated data) and are not encrypted, the encrypted data blocks CBi, for i between (h+1) and r, representing the payload, and the encrypted tag C_TAG. These blocks are added bit by bit to the current sequence S_j provided by the secondround logic unit 504 byXOR gate 510, however a value will only be provided online 512 when payload data blocks CB_h+1 to CB_r and encrypted tag C_TAG are being processed as no decryption is required for the header. For data blocks B₀ to B_h zeros will be provided online 512 toXOR gate 510, such that these data blocks pass throughgate 510 tooutput line 514 without being altered. The decrypted output ofXOR 510 is provided online 514, and represents the decrypted blocks of data B_i, which are also output directly on anoutput line 515.
• The output online 514 is also provided to asecond XOR gate 516 where it is added bit by bit to the output online 518 of atemporary tag register 520. The output ofXOR gate 516 is provided online 522 to the firstround logic unit 502. The result is provided to be stored in thetemporary tag register 520 online 524.
• As with the encryptingprocessing unit 300, the secondround logic unit 504 generates the encryption sequences S_j based on count values (nonce values N_j) from acounter 526 which receives an initialization value IV online 528 and a reset signal online 530. On each cycle, the counter adds 1, and outputs the value online 532.
• Units 304 and 504, and 302 and 502 respectively, are identical.
• Thesecret key 538, which is the same secret key as that used by processingunit 300, is provided to thekey schedule unit 506 online 540. The key schedule unit calculates a sub-key for each round of the first and second round logic units, and outputs these on bothline 542 to the firstround logic unit 502 andline 544 to the secondround logic unit 504. The order of generation of the sub-keys is the same for encrypting and decrypting.
• As for Figure 3, elements ofunit 500 receive control and synchronization signals (not shown).
• Once the tag has been calculated using all of the data blocks Bi, the result is provided to acomparison unit 534 online 518.Comparator 534 also receives the decrypted tag online 519. The received tag C_TAG is decrypted using, as for encryption, the initial sequence S₀, which is calculated by resettingcounter 526 to the IV, or in an alternative embodiment stored in a register when originally generated during processing of block B_h+1. Sequence S₀ is provided online 512 toXOR gate 510 to decrypt the tag, and the decrypted tag is then provided online 519 to the comparator.
• Thecomparator 534 provides a result online 536 indicating whether the tags match. If the tags match, the decrypted data blocks are deemed to be accurate, and can be processed as required. If the tags do not match however, this implies that the data has been corrupted, and therefore all the received data blocks and the tag are for example destroyed.
• The tag generation and ciphering (Figure 4) - or deciphering (Figure 5) - according to the first embodiment require r+2 cycles, compared to the 2(r+1)-h cycles required with the known solution of Figure 3.
• Figure 6 illustratesencryption processing unit 600 for encrypting data blocks B_i, and for generating a tag for data integrity according to a second embodiment. As will now be described, theencryption unit 600 is pipelined. Furthermore, a single pipeline is used for providing the functionality of both the CBC-MAC and counter mode round logic units.
• Data blocks B_i are received on aninput line 602 and added bit by bit, by anXOR gate 604, to theoutput 606 of atemporary tag register 608. The output ofXOR gate 604 is provided online 610 to a first input of a threeinput multiplexer 612.Multiplexer 612 also receives at a second input an input online 614 from acounter 616, which receives an initialization value online 618 and adds 1 on each payload data block cycle. The counter also receives a reset input online 620. At a third input,multiplexer 612 receives a signal online 622 from the output of the pipeline.
• Aselect input line 624 is provided for receiving an input signal SEL that selects which input ofmultiplexer 612 is selected for output onoutput line 626 of the multiplexer.
• Theoutput line 626 is connected to afirst stage 627 of the pipeline. The first stage comprises aregister 628 connected byline 630 to anunit 632 performing the SUBBYTES operations as described above withreferences 403 and 407 of figure 4.Register 628 acts as a buffer buffering data blocks received via themultiplexer 612. The SUBBYTES operation is particularly demanding on processing time and therefore it is preferable to arrange the pipeline with this function as the only function in a stage.
• The output ofunit 632 is provided to asecond stage 635 of the pipeline online 634. The second stage comprises aregister 636 for buffering the data output fromunit 632. The register outputs data values online 638 to aunit 640 performing theSHIFTROWS operations 404 and 408 of figure 4. The output ofunit 640 is provided online 642 to aunit 644 performing theMIXCOLUMNS operation 405 of figure 4.
• The output of theunit 644 is provided online 646 to aunit 648 performing the ADDROUNDKEY operation which receives sub-keys K₁ online 650 from akey schedule unit 652. Thekey schedule unit 652 generates sub-keys based on asecret key 654 provided online 656.
• The output of theunit 648 online 658 provides the output of thesecond stage 635 of the pipeline. This data value is selectively provided to thetemporary tag register 608 online 659, and to themultiplexer 612 online 622. The output is also provided to anXOR gate 660 which also receives the output of the temporary tag register online 606, and generates the encrypted tag C_TAG on itsoutput line 662. The second stage output is also provided to anXOR gate 664 which also receives online 666 the payload data block B_i to be encrypted and performs the XOR function bit by bit providing the encrypted data block CB_i (h+1≤i≤r) online 668.Line 666 is one of two outputs of afurther multiplexer 625, the single input of which is connected toline 602. The second output ofmultiplexer 625 outputs the header data blocks B₀ to B_h, which are not encrypted, online 669.
• Theprocessing unit 600 of figure 6 operates by alternatively processing first count values provided bycounter 616 for generating encryption sequences for encrypting the data blocks, and processing input data blocks for calculating the tag, the second stage being used for ending a round with either the nonce or the data associated with the current data block while the first stage starts another round with the data or the nonce. It should be noted that the same sub-key is needed for each round of the CBC-MAC algorithm for calculating the tag as for the equivalent round of the counter algorithm.
• The pipelined round is controlled to perform the same operations as disclosed in relation to figure 4. The header and payload of each packet is provided online 602. For each data block B₀ to B_h of the header (and associated data) which is only used to generate the tag and not encrypted, the pipeline operates only generating temporary tags TT_i.
• For each data block B₀ to B_h-1, the result of the combination of data block Bi with a previous temporary tag TT_i-1 (initialized with zero for combination with block B₀) is loaded into the pipeline viamultiplexer 612. This result is an intermediate tag IT_i. On the first round only theADDROUNDKEY unit 648 is used, and therefore the data block passes through theSUBBYTES unit 632 without being processed and is stored inregister 636. Then, the data block passes straight through theSHIFTROWS unit 640 and theMIXCOLUMNS unit 644 without being processed and is added bit by bit by theADDROUNDKEY unit 648 to the first sub-key K₀ generated by thekey schedule unit 652. The output (intermediate state IS₀) is returned online 622 to themultiplexer 612. On the second round and on subsequent rounds until the n-1^th round,SUBBYTES operation 632 is applied, as are all of the operations in the second stage of theprocessing unit 600. At the end of each round, the intermediate state IS₁ is passed back online 622 to be processed again. On the n^th and final round, all of the functions are applied except for theMIXCOLUMNS unit 644, and the output of the ADDROUNDKEY unit represents the temporary tag TT_i which is stored in theregister 608, ready to be combined with the next data block to be processed.
• When the B_h^th data block is to be processed, the encryption of the first payload data blocks B_h+1 can be prepared by encrypting nonce N_h+1 (=IV+1) to obtain the corresponding encryption sequence S_h+1 for the counter mode. Data block B_h is provided online 602, and combined atXOR gate 604 with the temporary tag TT_h-1 to produce the intermediate tag noted IT_h online 610. An IV value is also provided online 618, which is provided viacounter 616 as a nonce N_h+1 to themultiplexer 612 online 614. On a first half round period, the value IT_h is selected by themultiplexer 612 to be loaded into the pipeline and is transferred to register 636 (without processing). On a second half round period, the nonce N_h+1 online 614 is selected to be loaded into the pipeline and transferred to register 636 (without processing). During this second half round period, the first sub-key K₀ is applied to the value IT_h to produce the first intermediate state IS₀(IT_h) of the encryption of the value IT_h which is transferred, throughline 622 andmultiplexer 612, to register 628. On a third half round period, the intermediate state IS₀(IT_h) is processed by the first stage 627 (SUBBYTES operation) and the result is transferred to register 636 while the first sub-key K₀ is applied to the nonce N_h+1 to produce the first intermediate state IS₀(N_h+1) of the encryption of the nonce N_h+1 which is transferred, throughline 622 andmultiplexer 612, inregister 628.
• The same pipelined processing is reproduced for each intermediate state, the intermediate state IS_l(IT_h) ofrank 1 of the intermediate tag being processed by thefirst stage 627 while processing the previous intermediate value IS_l-1(N_h+1) of the nonce by thesecond stage 635 and the intermediate state IS₁(N_h+1) ofrank 1 of the nonce being processed by thefirst stage 627 while processing of the intermediate value IS_l(IT_h) of the same rank of the intermediate tag by thesecond stage 635. The current sub-key is buffered at the output ofunit 652 or at the input ofunit 648 to remain available for two successive computations by the second stage. Preferably, the sub-key K₁ is computed byunit 652 during the half period where the intermediate state IS_1-1 of the intermediate tag IT_h is processed by thefirst stage 627 and the intermediate state IS_1-1 of the nonce N_h+1 is processed by thesecond stage 635. On the 2(n+1)^th half period, first and second stages process intermediate states of rank n-1. The second stage produces the ciphered value of the intermediate tag IT_h which represents temporary tag TT_h of rank h. On the (2n+3)^th half period, the second stage produces the encryption sequence S_h+1 which is combined (gate 664) with the data block B_h+1 arriving on line 666 (trough themultiplexer 625 and line 602) to produce the first ciphered block CB_h+1. The data block B_h+1 is also combined (gate 604) with the temporary tag TT_h and the resulting intermediate tag IT_h+1 transferred in theregister 628 through themultiplexer 624. Further, the counter is incremented to produce the nonce N_h+2 online 614. Then, the operation explained above is reproduced for value IT_h+1 and nonce N_h+2, and so on until value IT_r-1 and nonce N_r.
• The first half round period of a cycle processing value ITi and nonce N_i+1 corresponds to the last half round period ((2n+3)^th) of a cycle processing value IT_i-1 and nonce N_i. Hence, n+1 round periods are enough to process both ciphered data block and corresponding temporary tag.
• On the (2n+3)^th half period of the n+1 cycle, as the Br^th is combined with the sequence S_r to produce the last cipher block CB_r, thecounter 616 is reinitialized (signal reset 620) to produce the nonce N₀ (preferably equal to the initialization value IV). Both stages are used, in a non-pipelined fashion as explained for the h+1 first blocks, to produce a Sequence S₀. This sequence is inputted with the content of theregister 608 toXOR gate 660 to produce the ciphered tag C_TAG.
• The tag generation and ciphering (Figure 6) according to the second embodiment require approximately r+3 cycles compared to the 2(r+1)-h cycles required with the known solution of Figure 3. Further, this second embodiment saves approximately the area of a round logic unit of the first embodiment.
• According to an alternative embodiment, the first sequence S₀ is computed in a pipelined way with the intermediate tag IT_h-1 and stored in a buffered input ofgate 660. An additional cycle is then saved.
• Other alternatives can be provided, for example the positioning of the registers organizing the pipeline can be varied. Further, even if the sharing of the operations presented in figure 6 is a preferred embodiment for time saving, the SHIFTROWS operation can be provided in the first stage (for example, before the SUBBYTES operation). Furthermore, a round could be sub-divided further, providing additional pipeline stages, or the stages could be arranged in a different manner.
• Pipelined decryption of data can be provided by similar circuitry to theencryption circuitry 600 used for encryption of data.
• Figure 7 illustrates one embodiment of pipelined decryption circuitry. In figure 7, the parts identical or similar with the circuitry of figure 6 have been labeled with the same reference numerals as figure 6, and these elements will not be described again. The first and second pipeline stages 627, 635 have not been shown in detail in figure 7, but are similar to those stages in figure 6.
• Circuit 700 comprises aninput line 702 on which are inputted the received header data blocks which have not been encrypted, the encrypted data blocks CB_i and the encrypted tag C_TAG. The circuit comprises anXOR gate 704 which decrypts the encrypted data blocks by performing an XOR function on them with the output of thesecond pipeline stage 635 received online 706, and in particular with the appropriate pre-calculated decryption sequence S_j. The output ofXOR gate 704 online 708 is provided to anXOR gate 604, and is also output online 710 providing the decrypted data block output of the circuit. This output is also provided vialine 712 to acomparator 714, which also receives the output of thetemporary tag register 608. Once the final tag has been calculated based on all of the received data blocks, it is stored in thetemporary tag register 608 before being compared with the received decrypted tag online 712. As before, if the tags match, the decrypted data blocks are deemed to be accurate, and can be processed as required. If the tags do not match however, this implies that the data has been corrupted, and therefore all of the received data block and the tag are for example destroyed.
• Operation of the circuit of figure 7 is essentially the same as the operation of thecircuit 600 of figure 6, except that the decryption sequence for each data block of the payload should be calculated in advance of the arrival of the encrypted data block so that the block may be decrypted before it is processed for calculating the tag value. Due to the nature of the XOR function, the decryption sequence for decrypting encrypted data blocks is in fact identical to the encryption sequence used to encrypt the data blocks.
• Figure 8 illustrates a variation to the pipelined encryption circuitry which provides a further reduction in the required hardware. Although not illustrated in figure 8, thefirst stage 827 of the pipeline is slightly different from thefirst stage 627 illustrated in figure 6 in that theSUBBYTES unit 632 ofstage 827 comprises only four S-boxes rather than the full sixteen S-boxes required for the 128-bit data blocks. Thus data is provided to the pipeline in sub-blocks each comprising 32 bits. For this, the circuit comprises three additional input registers 802, 804 and 806 for receiving the input data block online 610, the counter value online 614, and the previous output of the pipeline online 622 respectively. Each of these input registers receives 128 bits of data and divides these blocks into four sub-blocks of 32 bits for outputting sequentially to the pipeline.
• The pipeline operates in the same fashion as previously, except that instead of processing a block it processes a sub-block on each period. Themultiplexer 612 alternatively selects CBC-MAC data values and counter mode data values to be processed.
• The 32-bit output from the pipeline online 658 is provided to amultiplexer 808 which has two outputs onlines 810 and 812. The multiplexer is controlled to route the result to either a CBC-MAC output register 814 online 810 or a counter mode register 818 online 812. Once four processed sub-blocks have filled either of these output registers the 128-bit data value is provided to theoutput XOR gate 664, to thetemporary tag register 608, or to the return register 806 ready to be processed again by the pipeline.
• Thus embodiments have been described above that reduce the hardware requirements of CCM encryption and decryption circuitry, and increase the processing speed of this circuitry.
• Such encryption or decryption circuitry is of particular benefit in, for example, for wireless USB or in wireless LAN environments as described in the IEEE 802.11 protocol.
• By providing two processing cores, one for performing the round logic for CBC-MAC, and one for performing counter mode round logic, the payload of a packet can be processed twice as quickly as if a single conventional processor core were used, one for generating the tag, and again for encrypting the payload. Furthermore, hardware resources are minimized by using a common key schedule unit for generating sub-keys for both processor cores at the same time. This is made possible by arranging the processor cores to operate on blocks of data approximately synchronously.
• By pipelining a single processor core, it is possibly to process packet payloads for encryption and for generating a tag concurrently, and to perform these operations in approximately the time it would have taken to perform just one of these operations on a conventional single core architecture. The nature of the algorithms being performed do not make them immediately ideal for pipelining, particularly given that the output of one round is required before the next round may begin. By implementing a pipeline in which stages can be divided between the tag generation steps and the encryption generating steps, such that these processes can be performed concurrently, the throughput can be increased as well as reducing the required hardware. Again only a single key schedule block is required.
• Whilst a number of particular embodiments have been described above, there are many variations or alterations to these embodiments that may be applied. In particular, features of any of the described embodiments may be combined in alternative embodiments.
• Whilst the example of encrypting data and generating a tag based on a 128-bit key have been provided, the key, and the size of each data block, may be alternative lengths. The key schedule unit has been described as comprising S-boxes for generating sub-keys, however in alternative embodiments the sub-keys are a duplication of the secret key, or an alternative function thereof.
• The initialization value can be any value, and could for example comprise the source and/or destination address of the data. Furthermore, the unique count value can be generated using circuitry other than a counter.
• Whilst the encryption or decryption of data has been described as being performed by XOR gates, it could be performed using alternative logic. This equally applies to the XOR gate in the ADDROUNDKEY operation which could comprise alternative logic.
• Having thus described at least one illustrative embodiment of the invention, various alterations, modifications and improvements will readily occur to those skilled in the art. Such alterations, modifications and improvements are intended to be within the scope of the invention. Accordingly, the foregoing description is by way of example only and is not intended to be limiting. Further, the practical implementation of the present invention is in the ability of one with an ordinary skill in the art using common tools to obtain either a processor in wired logic, preferably programmable (FPGA). The invention is limited only as defined in the following claims and the equivalent thereto.

Claims (9)

1. Circuitry for encrypting at least a part of an input data (Bi) flow and generating a tag based on said input data flow with a same ciphering algorithm and a same key (K) comprising :

   a first ciphering branch arranged to encrypt said at least part of said input data;

   a second ciphering branch arranged to generate said tag; and

   a single key schedule unit (305) arranged to receive said key, to generate at least one sub-key based on said key and to provide said at least one sub-key to said first and second ciphering branches.
2. Circuitry according to claim 1, in which said first ciphering branch comprises:

   a first computation unit (304) arranged to successively receive first data values (N_j) and said at least one sub-key and to generate successive encryption sequences (S_j) with said ciphering algorithm; and

   a first logic unit (326) arranged to encrypt said at least part of said input data (B_i) flow based on said encryption sequences (N_j).
3. Circuitry according to claim 1 or 2, in which said second ciphering branch comprises:

   a second computation unit (302) arranged to generate successive temporary tags based on said ciphering algorithm and said at least one sub-key; and

   a second logic unit (308) comprising a first input (310) for successively receiving said temporary tags, a second input (306) for successively receiving said input data (B_i), and an output (313) connected to an input of said second computation unit.
4. Circuitry according to any preceding claim, further comprising a logic unit (330) for encrypting said tag with said first ciphering branch.
5. Circuitry for decrypting at least a part of an input encrypted data (CB_i) flow and recalculating a tag based on input data (B_i) with a same algorithm and a same key (K) comprising:

   a first deciphering branch arranged to decrypt said at least part of input data;

   a second ciphering branch arranged to generate said tag;

   a comparison unit to compare said generated tag with a received tag; and

   a single key schedule unit (305) arranged to receive said key, to generate at least one sub-key based on said key and to provide said at least one sub-key to said first and second branches.
6. Circuitry according to claim 5 in which said first deciphering branch comprises:

   a first computation unit (504) arranged to successively receive first data values (N_j) and said at least one sub-key and to generate successive decryption sequences (S_j) with said algorithm; and

   a first logic unit arranged to decrypt said at least part of said input data flow based on said decryption sequences (N_j).
7. Circuitry according to claim 5 or 6 in which said second ciphering branch comprises:

   a second computation unit (502) arranged to generate successive temporary tags based on said ciphering algorithm and said at least one sub-key; and

   a second logic unit (516) comprising a first input (518) for successively receiving said temporary tags, and a second input (514) for successively receiving said deciphered data (B_i).
8. The circuitry of any preceding claim, wherein said algorithm is the AES algorithm.
9. A wireless local area network element comprising circuitry for encrypting input data according to any of claims 1 to 4 and/or decrypting input data according to any of claims 5 to 8.

Images